CMBPR: Category-Aided Multi-Channel Bayesian Personalized Ranking for Short Video Recommendation

Abstract

Due to the viewing convenience for social media users' fragmented time, the short video has become a new carrier for users' network demands on information spread, news reading, social contact, entertainment, and leisure. Therefore, short video recommendation is one of the most important research topics in social media. Current short video recommendation algorithms mainly focus on detecting user's social attributes, developing cross-domain information and so on, few researchers combine video category information and multi-behavior information together. This paper proposes a content-based recommendation algorithm Category-aided Multi-channel Bayesian Personalized Ranking (CMBPR) for short video recommendation, which integrates users' rich preference information by considering the difference among both different video categories and different user interactions. The experimental results demonstrate the effectiveness of the CMBPR video recommendation algorithm, which achieves a significantly higher recommendation accuracy than the traditional video recommendation algorithms and solves the influence of the 'Long Tail' effect.